Tongue-Motion Classification Using Intraoral Electromyography for a Tongue-Computer Interface

We propose a human -- machine interface based on tongue motion for tetraplegics. The proposed system measures electromyography (EMG) signals in an oral cavity and estimate tongue motions. We measure four channels of intraoral EMG on the gum when a healthy subject moves the tongue in four directions (forward, backward, left, and right). For each EMG signal, we obtain the signal power for five frequency bands. We employ least-squares probabilistic classification (LSPC) to classify the tongue motions into these four motions and rest, and we obtained an 87.7 percent classification rate. Moreover, to attach a "click" function into the system, we measure the intraoral EMG signals caused by the clicking sound and those of other motions. Therefore, by embedding an EMG measurement system into an artificial tooth, we can realize a tongue -- computer interface that is not uncomfortable for users.